Skip to Main Content
In this paper, an on-line algorithm for automatic transcription of humming voice signal is proposed which is used in on-line Query-by-Humming systems. This algorithm consists of two stages: the first stage extracts the fundamental part of the humming voice by means of an octave detector. Constructing a state-space model for this mono-harmonic signal is extremely simpler than that of original signal. The second stage estimates and tracks the fundamental frequency of humming voice signal by means of an Extended Kalman Filter frequency tracker. In this way, each note and corresponding time duration will be identified. Because of on-line nature of the proposed technique, it does not suffer from any windowing problem and need no onset/offset detector for the notes and so, it is suitable for application in on-line Query-by-Humming systems.